PURA is a smart AI router that understands each pull request and reviews it with the right model based on your rules — within the budgets you set.
1-month free trial · BYO keys · We never store your code
Uncapped models on every PR add up fast. PURA puts the ceiling back: the right model per pull request, live spend, and fine-grained budgets per team, repo, and provider.
Routing a typo fix to a premium model burns budget on work that never needed it.
With no caps, one busy sprint can send your monthly review bill skyrocketing.
No idea which teams or repos drive AI cost — until the invoice lands.
No rollout project, no new workflow. Engineers get fast reviews; leaders get predictable spend — right inside GitHub.
Install in a couple of clicks, connect your Claude, OpenAI, or Gemini key, and pick the models you trust.
Tell PURA which changes need a premium review and which can go lean — in plain English. Then set daily, weekly, and monthly caps.
/pura reviewDrop /pura review on any PR. PURA sizes the change, picks the right model, and posts one clear review — summary, ranked findings, ready-to-apply fixes.
Every review is tracked live against your caps. Reach a cap and PURA pauses that agent and skips the PR — so you never spend a dollar over.
Just write your routing rules in plain English in a .pura/PURA.md file. PURA reads each PR and your rules, then picks the right model — a top model for risky changes, a cheap one for routine work.
# Just write it in plain English.
If the PR changes src/security.ts
or anything under src/payments/,
review it with gpt-5.5.
For docs, README, and dependency
bumps, use gpt-5.4-nano.
Everything else: gpt-5.4-mini.The number you commit to finance is the number you'll pay — it physically cannot be exceeded, while every team keeps moving fast inside its own lane.
One top-line cap per agent — the number you give finance.
Give each squad its own envelope to spend.
Cap noisy, high-traffic, or high-risk repos.
Stop any single author from running away.
A hard ceiling on what one PR can cost.
Split spend across OpenAI, Anthropic & Google.
Hit any budget and PURA falls over to a backup provider or pauses reviews — it never sends you a surprise invoice.
Engineers get fast, focused reviews. Leaders get predictable spend, security they can approve, and keys they own.
Daily, weekly, and monthly caps per team, repo, and provider — with live spend and leaderboards in the dashboard. The bill is set before the work starts.
Bring your own Claude, OpenAI, or Gemini account. Inference is billed by your provider with zero markup, and keys are encrypted the moment you save them.
We never store your code, secrets are redacted before any model sees them, and admin controls cover models, keys, and budgets.
No per-seat fees, no hidden charges. A straightforward fee based on the number of reviews.
For a team getting AI review under control.
For growing orgs reviewing across many repos.
For larger orgs that need depth and SLAs.
For security, scale, and procurement needs.
You bring your own AI provider keys, so inference is billed by your provider, never marked up by PURA. Prices exclude VAT.
Bring the questions your CTO, security lead, or finance partner will ask. In 30 minutes we'll show the product, the controls, and the ROI.
Everything teams ask before turning on AI code review.
PURA charges a flat platform fee per plan, metered by the number of reviews, with no per-seat fees and no hidden charges. You bring your own model API keys and pay your AI provider directly for inference — PURA never marks it up.
Yes. PURA is a GitHub App. Comment /pura review on any pull request and PURA posts one clear review — a summary, findings ranked by severity, and ready-to-apply fixes — right in GitHub.
Yes. PURA is bring-your-own-key: connect your own OpenAI, Anthropic, or Google (Gemini) account and inference is billed directly by that provider with zero markup. Your keys are encrypted at rest and never shown again after you save them.
PURA distributes fine-grained budgets across every level — per agent, team, repo, provider, even per pull request — on daily, weekly, and monthly windows. When a budget is reached, PURA pauses reviews instead of overspending, so your bill stays within the numbers you set. A spend dashboard shows exactly where every dollar goes.
You decide. Write routing rules in plain English in a .pura/PURA.md file so PURA sends sensitive changes (like payments or auth) to premium models and routine changes (like docs or dependency bumps) to cheaper ones.
No. Code is processed transiently to generate a review and is not retained afterward, and detected secrets are redacted before anything is sent to the model. PURA does not use your source code to train AI models.
No. PURA's reviews are advisory and assist your team. You stay in control of every merge and release decision.